More turbulence ahead?

THE local semiconductor industry faced trying times in 2025 stemming from policy uncertainty, tariff woes, global market volatility and the much-debated artificial intelligence (AI) bubble.

How will the sector fare this year?

Peter Lim Tze Cheng, a fund manager who closely follows the local semiconductor sector, sees 2025 as “a year of recovery” for the industry.

“Though we saw sentiment shift several times, driven by global market volatility, profit-taking at year-end, and ongoing uncertainty about policies and tariffs, the recovery of the sector can be clearly seen,” he tells StarBiz 7.

The artificial intelligence (AI) frenzy dominated discussions in the semiconductor sector for most of last year.

By late 2025, concerns about an AI bubble grew. Companies that invested heavily in AI infrastructure may struggle to meet profit expectations.

Big tech firms continue to spend billions on chips and data centres, but investors are watching for signs that AI demand might be slowing.

A sharp drop in revenue growth could signal trouble for the overall AI investment trend.

Hence 2026 is seen as a crucial year for testing AI’s profitability, based on the orderbooks of major semiconductor companies like Taiwan Semiconductor Manufacturing Co Ltd, Broadcom Inc, and Nvidia Corp.

Danish Lim Zhi Lin, a senior analyst at Phillip Nova, the futures trading arm of Phillip Capital, calls this risk “the global AI capex deceleration,” referring to the chance of a sharp drop in investments that could harm some semiconductor companies, including those in Malaysia.

“Malaysia’s main participation in the AI silicon value chain is via back-end assembly, testing and packaging and adjacent equipment, automation and inspection capabilities.

“These areas can benefit as AI shifts towards higher complexity per chip,” he says.

Automated test equipment (ATE) firm QES Group Bhd’s managing director Chew Ne Weng says the risk of an AI bubble bursting is real, though it is “not too soon and probably can happen from 2027 onwards”.

He adds that AI monetisation might not happen this year, with significant returns expected only by 2027 or 2028.

A global AI downturn could also negatively impact the broader electrical and electronics (E&E) market.

Thomas Yong of Fortress Capital sees a promising future for Malaysian semiconductor companies that have shifted from consumer electronics to AI infrastructure.

He also expects higher demand for advanced ATE due to AI growth.

On the other hand, companies that haven’t or can’t make this shift face a tougher outlook.

Many in the local ecosystem are still tied to non-AI legacy sectors, which could struggle to recover in 2026, he notes.

Another headwind is rising operational costs. “This is driven by the multi-tier levy system and minimum wage hikes, at a time when global customers are demanding more complex, capital intensive work.

“Things are further complicated by the strengthening of the ringgit,” says Yong.

Semiconductor companies in consumer-facing markets, like smartphones and personal computers, are expected to lag behind.

A notable trend is the recent rise in memory prices, which might lead to fewer smartphone and PC shipments due to increased costs.

There are also risks of sector-specific tariffs on semiconductors, which might push investments away from Malaysia, Zhi Lin says.

Last year, during heightened tensions, US President Donald Trump suggested tariffs of up to 300% on semiconductors.

These tariffs have not been enforced and are under review in the US Section 232 investigation.

For now, Malaysian semiconductor exports to the US avoid the general 19% reciprocal tariff but the future remains uncertain.

“The current situation remains fluid, with the possibility of further adjustments amid ongoing negotiations and potential exemptions. Final rates will depend on negotiation results and Trump’s decision,” Zhi Lin says.

Chew says if Section 232 turns negative, “it will have adverse impact on most semiconductor sectors”.

Tze Cheng, a chief research officer at Trident Analytics Sdn Bhd, however, brushes off tariff concerns, opining that it is “more of a scare tactic” by Trump to push companies to bring manufacturing facilities back to the US.

“Unlike sectors like steel or automotive, many of the largest suppliers exporting semiconductor products into the US are actually US corporations themselves.

“Hence, it does not really make sense for him to impose tariffs on semiconductors, because it defeats the purpose of what he is trying to achieve, which is to ensure the US benefits from granting market access to other countries,” he says.

Tze Cheng believes the local semiconductor market will see a three-year boom cycle starting from this year, driven by delayed positive spillover from global AI demand and a recovery in non-AI segments after two years of cyclical weakness.

The listed semiconductor companies on Bursa Malaysia perform poorly last year.

The Bursa Malaysia Technology Index dropped nearly 15% last year.

According to Tze Cheng, the weak performance was due to profit-taking.

“The sector had performed reasonably well, but as the year ended, investors locked in their profits for 2025, which caused the tech index to weaken,” he explains.

Looking ahead, Tze Cheng anticipates strong double-digit growth across the sector.

“From my conversations with businesses, there is a broad agreement that business is coming in, in a fairly significant way,” he says.

Chew attests to this, noting the group is getting “much more sales inquiries as its semiconductor clientele starts to gradually ramp up production”.

Tze Cheng explains that the current AI wave is not a bubble like past ones, such as the Dutch tulip craze or the dot-com meltdown.

He highlights that AI is backed by real products, services, and growing demand.

He notes that every technology cycle needs a “killer app”; a must-have application that drives widespread use and justifies big investments.

For example, social media and ride-hailing services were the killer apps for 4G, making the upgrade from 3G essential.

“5G adoption has lagged because there are no applications that cannot already be supported by 4G. Similarly for AI, the killer app has yet to emerge.

“So far, LLMs are the most visible use cases, but these represent only the surface level of AI.”